update pytorchcv.

diceryas · Jan 10, 2019 · 3705979 · 3705979
1 parent 141c0d2
commit 3705979
Show file tree

Hide file tree

Showing 346 changed files with 6,974 additions and 27,753 deletions.
diff --git a/BUGS.md b/BUGS.md
@@ -0,0 +1,4 @@
+### Bugs with PyTorchCV
+
+- Backbone with pretrained parameters is initialized, too.
+
diff --git a/README.md b/README.md
@@ -1,23 +1,18 @@
 # PyTorchCV: A PyTorch-Based Framework for Deep Learning in Computer Vision
 ```
 @misc{CV2018,
-  author =       {Donny You & SIGAI (www.sigai.cn)},
-  howpublished = {\url{https://github.com/CVBox/PyTorchCV}},
+  author =       {Donny You ([email protected])},
+  howpublished = {\url{https://github.com/youansheng/PyTorchCV}},
   year =         {2018}
 }
 ```
 
-&ensp;&ensp;This repository provides source code for some deep learning based cv problems. We'll do our best to keep this repository up to date.  If you do find a problem about this repository, please raise it as an issue. We will fix it immediately.
-
-&ensp;&ensp;SIGAI, aiming to develop AI framework for Everyone. Please visit [www.sigai.cn](http://www.sigai.cn/)
-<div align="center">
-<img src="docs/_static/sigai.png" width="300px"/>
-</div>
+This repository provides source code for some deep learning based cv problems. We'll do our best to keep this repository up to date.  If you do find a problem about this repository, please raise it as an issue. We will fix it immediately.
 
 
 ## Implemented Papers
 
-- [Image Classification](https://github.com/CVBox/PyTorchCV/tree/master/methods/cls)
+- [Image Classification](https://github.com/youansheng/PyTorchCV/tree/master/methods/cls)
     - VGG: Very Deep Convolutional Networks for Large-Scale Image Recognition
     - ResNet: Deep Residual Learning for Image Recognition
     - DenseNet: Densely Connected Convolutional Networks
@@ -27,22 +22,22 @@
     - ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices
     - ShuffleNet V2: Practical Guidelines for Ecient CNN Architecture Design
 
-- [Semantic Segmentation](https://github.com/CVBox/PyTorchCV/tree/master/methods/seg)
+- [Semantic Segmentation](https://github.com/youansheng/PyTorchCV/tree/master/methods/seg)
     - DeepLabV3: Rethinking Atrous Convolution for Semantic Image Segmentation
     - PSPNet: Pyramid Scene Parsing Network
     - DenseASPP: DenseASPP for Semantic Segmentation in Street Scenes
 
-- [Object Detection](https://github.com/CVBox/PyTorchCV/tree/master/methods/det)
+- [Object Detection](https://github.com/youansheng/PyTorchCV/tree/master/methods/det)
     - SSD: Single Shot MultiBox Detector
     - Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
     - YOLOv3: An Incremental Improvement
     - FPN: Feature Pyramid Networks for Object Detection
 
-- [Pose Estimation](https://github.com/CVBox/PyTorchCV/tree/master/methods/pose)
+- [Pose Estimation](https://github.com/youansheng/PyTorchCV/tree/master/methods/pose)
     - CPM: Convolutional Pose Machines
     - OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
 
-- [Instance Segmentation](https://github.com/CVBox/PyTorchCV/tree/master/methods/seg)
+- [Instance Segmentation](https://github.com/youansheng/PyTorchCV/tree/master/methods/seg)
     - Mask R-CNN
 
 
@@ -57,24 +52,24 @@
 | Model | Backbone | Training data  | Testing data | mIOU | Pixel Acc | Setting |
 |--------|:---------:|:------:|:------:|:------:|:------:|:------:|
 | [PSPNet Origin](https://github.com/hszhao/PSPNet) | 3x3-ResNet101 | ADE20K train | ADE20K val | 41.96 | 80.64 | - |
-| [PSPNet Ours](https://drive.google.com/open?id=1Q6oYBpq9Y53z_CJz7Km9BaiSVJjcHP4h) | [7x7-ResNet101](https://drive.google.com/open?id=1ROewKyaGPynox_-a50wHkSv1-0jYWyvc) | ADE20K train | ADE20K val | 44.18 | 80.91 | [PSPNet](https://github.com/CVBox/PyTorchCV/blob/master/hypes/seg/ade20k/fs_pspnet_ade20k_seg.json) |
+| [PSPNet Ours](https://drive.google.com/open?id=1Q6oYBpq9Y53z_CJz7Km9BaiSVJjcHP4h) | [7x7-ResNet101](https://drive.google.com/open?id=1ROewKyaGPynox_-a50wHkSv1-0jYWyvc) | ADE20K train | ADE20K val | 44.18 | 80.91 | [PSPNet](https://github.com/youansheng/PyTorchCV/blob/master/hypes/seg/ade20k/fs_pspnet_ade20k_seg.json) |
 
 #### Object Detection
 - SSD: Single Shot MultiBox Detector
 
 | Model | Backbone | Training data  | Testing data | mAP | FPS  | Setting |
 |--------|:-------:|:---------:|:------:|:------:|:------:|:------:|
 | [SSD-300 Origin](https://github.com/weiliu89/caffe/tree/ssd) | VGG16 | VOC07+12 trainval | VOC07 test | 0.772 | - | - |
-| [SSD-300 Ours](https://drive.google.com/open?id=15J5blVyZq7lqCePh-Q8S2pxim3-f_8LP) | [VGG16](https://drive.google.com/open?id=1nM0UwmqR4lIHzmRWvs71jfP_gAekjuKy) | VOC07+12 trainval | VOC07 test | 0.786 | - | [SSD300](https://github.com/CVBox/PyTorchCV/blob/master/hypes/det/voc/ssd_vgg300_voc_det.json) |
+| [SSD-300 Ours](https://drive.google.com/open?id=15J5blVyZq7lqCePh-Q8S2pxim3-f_8LP) | [VGG16](https://drive.google.com/open?id=1nM0UwmqR4lIHzmRWvs71jfP_gAekjuKy) | VOC07+12 trainval | VOC07 test | 0.786 | - | [SSD300](https://github.com/youansheng/PyTorchCV/blob/master/hypes/det/voc/ssd_vgg300_voc_det.json) |
 | [SSD-512 Origin](https://github.com/weiliu89/caffe/tree/ssd) | VGG16 | VOC07+12 trainval | VOC07 test | 0.798 | - | - |
-| [SSD-512 Ours](https://drive.google.com/open?id=1RF5gnqfiyz-EcSFU1OSK7tNuX_VRObVW) | [VGG16](https://drive.google.com/open?id=1nM0UwmqR4lIHzmRWvs71jfP_gAekjuKy) | VOC07+12 trainval | VOC07 test | 0.808 | - | [SSD512](https://github.com/CVBox/PyTorchCV/blob/master/hypes/det/voc/ssd_vgg512_voc_det.json) |
+| [SSD-512 Ours](https://drive.google.com/open?id=1RF5gnqfiyz-EcSFU1OSK7tNuX_VRObVW) | [VGG16](https://drive.google.com/open?id=1nM0UwmqR4lIHzmRWvs71jfP_gAekjuKy) | VOC07+12 trainval | VOC07 test | 0.808 | - | [SSD512](https://github.com/youansheng/PyTorchCV/blob/master/hypes/det/voc/ssd_vgg512_voc_det.json) |
 
 - Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
 
 | Model | Backbone |  Training data  | Testing data | mAP | FPS  | Setting |
 |--------|:--------:|:---------:|:------:|:------:|:------:|:------:|
 | [Faster R-CNN Origin](https://github.com/rbgirshick/py-faster-rcnn) | VGG16 | VOC07 trainval | VOC07 test | 0.699 | - | - |
-| [Faster R-CNN Ours](https://drive.google.com/open?id=15SfklRiI1McVWEq9EAceznK-9sxXSQR4) | [VGG16](https://drive.google.com/open?id=1ZL9SS9KRzsDQhMe8kyPQ1LHA60wx_Vcj) | VOC07 trainval | VOC07 test | 0.706 | - | [Faster R-CNN](https://github.com/CVBox/PyTorchCV/blob/master/hypes/det/voc/fr_vgg16_voc_det.json) |
+| [Faster R-CNN Ours](https://drive.google.com/open?id=15SfklRiI1McVWEq9EAceznK-9sxXSQR4) | [VGG16](https://drive.google.com/open?id=1ZL9SS9KRzsDQhMe8kyPQ1LHA60wx_Vcj) | VOC07 trainval | VOC07 test | 0.706 | - | [Faster R-CNN](https://github.com/youansheng/PyTorchCV/blob/master/hypes/det/voc/fr_vgg16_voc_det.json) |
 
 - YOLOv3: An Incremental Improvement
 

diff --git a/datasets/cls/cifar/cifar_cls_generator.py b/datasets/cls/cifar/cifar_cls_generator.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 
 
 from __future__ import print_function

diff --git a/datasets/cls/cifar/cifar_cls_generator.sh b/datasets/cls/cifar/cifar_cls_generator.sh
@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Generate train & val data.
 
 

diff --git a/datasets/cls/fc_data_loader.py b/datasets/cls/fc_data_loader.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Data loader for Image Classification.
 
 
@@ -10,9 +10,9 @@
 
 import json
 import os
-
 import torch.utils.data as data
 
+from extensions.parallel.data_container import DataContainer
 from utils.helpers.image_helper import ImageHelper
 
 
@@ -37,7 +37,10 @@ def __getitem__(self, index):
         if self.img_transform is not None:
             img = self.img_transform(img)
 
-        return img, label
+        return dict(
+            img=DataContainer(img, stack=True),
+            label=DataContainer(label, stack=True),
+        )
 
     def __len__(self):
 

diff --git a/datasets/cls/flower/flower_cls_generator.py b/datasets/cls/flower/flower_cls_generator.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # KITTI det data generator.
 
 

diff --git a/datasets/cls/flower/flower_cls_generator.sh b/datasets/cls/flower/flower_cls_generator.sh
@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Generate train & val data.
 
 

diff --git a/datasets/cls/imagenet/imagenet_cls_generator.py b/datasets/cls/imagenet/imagenet_cls_generator.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # KITTI det data generator.
 
 

diff --git a/datasets/cls/imagenet/imagenet_cls_generator.sh b/datasets/cls/imagenet/imagenet_cls_generator.sh
@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Generate train & val data.
 
 

diff --git a/datasets/cls_data_loader.py b/datasets/cls_data_loader.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Class for the Pose Data Loader.
 
 
@@ -15,7 +15,7 @@
 import datasets.tools.cv2_aug_transforms as cv2_aug_trans
 import datasets.tools.transforms as trans
 from datasets.cls.fc_data_loader import FCDataLoader
-from datasets.tools.collate_functions import CollateFunctions
+from datasets.tools.collate import collate
 from utils.tools.logger import Logger as Log
 
 
@@ -54,9 +54,9 @@ def get_trainloader(self):
                              img_transform=self.img_transform, configer=self.configer),
                 batch_size=self.configer.get('train', 'batch_size'), shuffle=True,
                 num_workers=self.configer.get('data', 'workers'), pin_memory=True,
-                collate_fn=lambda *args: CollateFunctions.our_collate(
-                    *args, data_keys=['img', 'label'],
-                    trans_dict=self.configer.get('train', 'data_transformer')
+                drop_last=self.configer.get('data', 'drop_last'),
+                collate_fn=lambda *args: collate(
+                    *args, trans_dict=self.configer.get('train', 'data_transformer')
                 )
             )
 
@@ -74,9 +74,9 @@ def get_valloader(self):
                              img_transform=self.img_transform, configer=self.configer),
                 batch_size=self.configer.get('val', 'batch_size'), shuffle=False,
                 num_workers=self.configer.get('data', 'workers'), pin_memory=True,
-                collate_fn=lambda *args: CollateFunctions.our_collate(
-                    *args, data_keys=['img', 'label'],
-                    trans_dict=self.configer.get('val', 'data_transformer')
+                drop_last=self.configer.get('data', 'drop_last'),
+                collate_fn=lambda *args: collate(
+                    *args, trans_dict=self.configer.get('val', 'data_transformer')
                 )
             )
 

diff --git a/datasets/det/aa_data_loader.py b/datasets/det/aa_data_loader.py
diff --git a/datasets/det/coco/coco_det_generator.py b/datasets/det/coco/coco_det_generator.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # COCO det data generator.
 
 

diff --git a/datasets/det/coco/coco_det_generator.sh b/datasets/det/coco/coco_det_generator.sh
@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Generate train & val data.
 
 

diff --git a/datasets/det/fashion/fashion_det_generator.py b/datasets/det/fashion/fashion_det_generator.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Fashion det data generator.
 
 

diff --git a/datasets/det/fashion/fashion_det_generator.sh b/datasets/det/fashion/fashion_det_generator.sh
@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Generate train & val data.
 
 

diff --git a/datasets/det/fr_data_loader.py b/datasets/det/fr_data_loader.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Faster R-CNN data loader
 
 
@@ -13,6 +13,7 @@
 import numpy as np
 import torch.utils.data as data
 
+from extensions.parallel.data_container import DataContainer
 from utils.helpers.json_helper import JsonHelper
 from utils.helpers.image_helper import ImageHelper
 from utils.tools.logger import Logger as Log
@@ -39,17 +40,25 @@ def __getitem__(self, index):
         if self.aug_transform is not None:
             img, bboxes, labels = self.aug_transform(img, bboxes=bboxes, labels=labels)
 
+        img_scale = ImageHelper.get_size(img)[0] / img_size[0]
+
         labels = torch.from_numpy(labels).long()
         bboxes = torch.from_numpy(bboxes).float()
 
-        scale1 = 600 / min(img_size)
-        scale2 = 1000 / max(img_size)
-        scale = min(scale1, scale2)
-
+        meta = dict(
+            ori_img_size=img_size,
+            aug_img_size=ImageHelper.get_size(img),
+            img_scale=img_scale,
+        )
         if self.img_transform is not None:
             img = self.img_transform(img)
 
-        return img, scale, bboxes, labels
+        return dict(
+            img=DataContainer(img, stack=True),
+            bboxes=DataContainer(bboxes, stack=False),
+            labels=DataContainer(labels, stack=False),
+            meta=DataContainer(meta, stack=False, cpu_only=True)
+        )
 
     def __len__(self):
 

diff --git a/datasets/det/kitti/kitti_det_generator.py b/datasets/det/kitti/kitti_det_generator.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # KITTI det data generator.
 
 

diff --git a/datasets/det/kitti/kitti_det_generator.sh b/datasets/det/kitti/kitti_det_generator.sh
@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 # -*- coding:utf-8 -*-
-# Author: Donny You(donnyyou@163.com)
+# Author: Donny You(youansheng@gmail.com)
 # Generate train & val data.
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1,4 @@
		### Bugs with PyTorchCV

		- Backbone with pretrained parameters is initialized, too.